Exploring the utility of GRE-MRI signal compartments for temporal susceptibility mapping and investigation of neural microstructure

Abstract

Magnetic resonance imaging (MRI) has played a key role in our understanding of the brain’s anatomy and physiology. In particular, gradient recalled echo magnetic resonance imaging (GRE-MRI) at ultra-high field holds great promise for new contrast mechanisms to examine brain structure non-invasively. Multi-echo GRE-MRI is affected by signal compartments which may inform structural characterization. A number of studies have adopted the three water-pool compartment model to study white matter brain regions by associating individual compartments with myelin, axonal and extracellular water. Many key questions, however, remain unanswered in the context of GRE-MRI signal compartmentalization. First, the number and identifiability of GRE-MRI signal compartments has not been fully explored. We examined these issues in the human brain using a data driven approach, as detailed in Ch. 1. Multiple echo time GRE-MRI data were acquired in five healthy participants, each brain was segmented into anatomical regions (substantia nigra, caudate, insula, putamen, thalamus, fornix, internal capsule, corpus callosum and cerebrospinal fluid) and the temporal signal fitted with models with one to six signal compartments. With the use of information criteria and cluster analysis methods we ascertained the number of distinct signal compartments within each region and established differences in their respective frequency shifts between the brain regions studied. We identified five dominant signal compartments; these contributed to the local signal frequency of each brain region differently. Maps of compartment volume fractions, resolved by fixing the respective compartment frequency shifts in each voxel, corresponded with commonly observed tissue properties.Second, the influence of the scanner field strength on the temporal evolution of the multiecho GRE-MRI signal is not known. Subsequently, the influence of scanner field strength on GRE-MRI signal compartments also remains uncharacterized. In Ch. 3, we evaluated variations in the signal frequency shifts due to changes in field strength within the putamen, CSF, and corpus callosum (representatives of gray matter, CSF, and white mater regions respectively). Multiple echo time GRE-MRI data at 3T and 7T were acquired in six healthy participants, and temporal frequency shift profiles were generated via the quantitative susceptibility pipeline. From a qualitative lens, we observed unique echo-time dependent frequency shift profiles for each brain region to broadly correspond between 3T and 7T measurements. Furthermore, inter-participant variability in frequency shifts were higher in 3T measurements than at 7T. In general, signal compartment frequency shifts correspond well between 3T and 7T data, and standard error estimates indicate an improved quality-of- fit within the putamen and CSF, when compared to the corpus callosum. Compartment frequency shifts mapped using multi-echo GRE-MRI signal compartment models may thus provide new insights into tissue composition and structure, holding particular value in generating biomarkers for neurodegenerative diseases and psychiatric disorders.